On Mon, Aug 29, 2016 at 03:08:44PM +0530, Srikar Dronamraju wrote:

quoted

Patch "mm: vmscan: Begin reclaiming pages on a per-node basis" started
thinking of reclaim in terms of nodes but kswapd is still zone-centric. This
patch gets rid of many of the node-based versus zone-based decisions.

o A node is considered balanced when any eligible lower zone is balanced.
  This eliminates one class of age-inversion problem because we avoid
  reclaiming a newer page just because it's in the wrong zone
o pgdat_balanced disappears because we now only care about one zone being
  balanced.
o Some anomalies related to writeback and congestion tracking being based on
  zones disappear.
o kswapd no longer has to take care to reclaim zones in the reverse order
  that the page allocator uses.
o Most importantly of all, reclaim from node 0 with multiple zones will
  have similar aging and reclaiming characteristics as every
  other node.

Signed-off-by: Mel Gorman <redacted>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <redacted>

This patch seems to hurt FA_DUMP functionality. This behaviour is not
seen on v4.7 but only after this patch.

So when a kernel on a multinode machine with memblock_reserve() such
that most of the nodes have zero available memory, kswapd seems to be
consuming 100% of the time.

Why is FA_DUMP specifically the trigger? If the nodes have zero available
memory then is the zone_populated() check failing when FA_DUMP is enabled? If
so, that would both allow kswapd to wake and stay awake.

-- 
Mel Gorman
SUSE Labs